@@ -483,8 +483,10 @@ static void atc_advance_work(struct at_dma_chan *atchan)
dev_vdbg(chan2dev(&atchan->chan_common), "advance_work\n");
spin_lock_irqsave(&atchan->lock, flags);
- if (atc_chan_is_enabled(atchan) || list_empty(&atchan->active_list))
- return spin_unlock_irqrestore(&atchan->lock, flags);
+ if (atc_chan_is_enabled(atchan) || list_empty(&atchan->active_list)) {
+ spin_unlock_irqrestore(&atchan->lock, flags);
+ return;
+ }
desc = atc_first_active(atchan);
/* Remove the transfer node from the active list. */
@@ -1477,8 +1479,10 @@ static void atc_issue_pending(struct dma_chan *chan)
dev_vdbg(chan2dev(chan), "issue_pending\n");
spin_lock_irqsave(&atchan->lock, flags);
- if (atc_chan_is_enabled(atchan) || list_empty(&atchan->queue))
- return spin_unlock_irqrestore(&atchan->lock, flags);
+ if (atc_chan_is_enabled(atchan) || list_empty(&atchan->queue)) {
+ spin_unlock_irqrestore(&atchan->lock, flags);
+ return;
+ }
desc = atc_first_queued(atchan);
list_move_tail(&desc->desc_node, &atchan->active_list);
v5.10.234-rt127-rc1 stable review patch. If anyone has any objections, please let me know. ----------- The problem described here is specific to v5.10-rt. The functions atc_advance_work() and atc_issue_pending(), defined at drivers/dma/at_hdmac.c, both have a similar statement: return spin_unlock_irqrestore(&atchan->lock, flags); That results in a macro expansion error during build. This problem is unique to v5.10-rt because of the way spin_unlock_irqrestore() is defined as a macro at include/linux/spinlock_rt.h. Newer versions of the PREEMPT_RT patch (v5.15-rt and newer) have spin_unlock_irqrestore() defined as a function. Kernels such as v4.19-rt and older are also not affected because they don't have the two commits below, which implement the statements that trigger the compiler error: 1582cc3b4805 dmaengine: at_hdmac: Fix concurrency problems by removing atc_complete_all() 7078e935b410 dmaengine: at_hdmac: Fix premature completion of desc in issue_pending It makes more sense, at the current point in v5.10-rt life cycle, decoupling the return and spin_unlock_irqrestore() statements instead of backporting the new definition of spin_unlock_irqrestore() and related bits. Signed-off-by: Luis Claudio R. Goncalves <lgoncalv@redhat.com> --- drivers/dma/at_hdmac.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-)